DeepSeek API Pricing vs OpenAI: Where We Fit
Comparing token prices only works if you use your own workload. This page gives our exact rates, a formula, three worked examples and the cases where another provider is the better pick.
Updated

Key points
- Our rates are $0.15 per 1M input tokens and $0.50 per 1M output tokens, with no subscription.
- Compare with cost = input / 1M x input price + output / 1M x output price, using the other provider's current price page.
- This service is neither the DeepSeek API nor OpenAI; it runs one independent open-weight model.
- OpenAI is the better fit if you need images, audio, embeddings or fine-tuning, which we do not offer.
What this comparison can and cannot say
If you are weighing deepseek api pricing vs openai, you will find plenty of tables online, and most of them age quickly because providers change prices and release new models. I will not quote any other company's rates here. Instead this article gives you our exact numbers, the formula for comparing any two price lists, and worked examples so you can plug in figures you check yourself on each provider's current pricing page.
Two clarifications up front. This site is not the official DeepSeek API, and it is not OpenAI. It is an independent, OpenAI-compatible service that runs one open-weight model, tuned to answer without content refusals for lawful adult use. The model ID is uncensored. It is not a router across vendors. So the real question for most visitors is not "which of two big names is cheaper" but "what does my workload cost here, and does this model do the job".
The general DeepSeek side of the comparison is covered in the DeepSeek Uncensored API pricing guide on this site. This article stays on method and our numbers.
Our price list in one table
| Item | Value |
|---|---|
| Input tokens | $0.15 per 1M |
| Output tokens | $0.50 per 1M |
| Billing | Pay as you go from prepaid credit, no subscription |
| Failed or refused requests | Not charged |
| Credit expiry | Never expires |
| Top-ups | Any whole amount from $20 to $500 |
| Top-up bonus | +5% from $50, +10% from $100, +15% from $200 |
| Payment | Crypto (USDT on TRC20, USDC on Base) |
| Trial | $0.50 credit, one click after sign-up, 7 days, no card |
| Context window | 100,000 tokens, prompt plus completion |
The bonuses are easy to turn into money. A $100 top-up gives +10%, so $10 of extra credit. A $200 top-up gives +15%, which is $30 extra, for $230 total. A $50 top-up gives +5%, which is $2.50. Details live on the pricing page.
Output costs a bit more than three times as much as input ($0.50 against $0.15). That ratio decides which workloads are cheap here, which the examples below show.
The formula
Any chat-completion price list reduces to two numbers: a price for input tokens and a price for output tokens, usually quoted per million. Your monthly bill is:
cost = (input tokens / 1,000,000) x input price + (output tokens / 1,000,000) x output price
To compare two providers, do not look at one headline number. Take your own typical request, count its input and output tokens, and run the formula once for each price list. Check three details on the other provider's page as you do it: whether cached or repeated input is priced differently, whether there are different tiers for different models, and whether there are minimums or monthly fees. Those differences can outweigh a small gap in the headline rates, and I cannot tell you them for other companies. Check their current pricing.
The simplest honest way to get your own token counts is to run a few hundred real requests and read the usage field. Here is a small script that does the cost math for this endpoint after each call:
from openai import OpenAI
PRICE_IN = 0.15 # dollars per 1M input tokens
PRICE_OUT = 0.50 # dollars per 1M output tokens
client = OpenAI(base_url="https://api.uncensoreddeepseek.com/v1", api_key="YOUR_KEY")
resp = client.chat.completions.create(
model="uncensored",
messages=[{"role": "user", "content": "Summarize the plot of Hamlet in 120 words."}],
)
u = resp.usage
cost = u.prompt_tokens / 1_000_000 * PRICE_IN + u.completion_tokens / 1_000_000 * PRICE_OUT
print(u.prompt_tokens, u.completion_tokens, round(cost, 6))
Log those three numbers for a sample of real traffic, then replace the two constants with another provider's prices to see the other side.
Three workloads, worked out
A support chatbot
100,000 requests per month, each with 1,500 input tokens (instructions plus history) and 300 output tokens.
- Input: 100,000 x 1,500 = 150M tokens x $0.15 = $22.50
- Output: 100,000 x 300 = 30M tokens x $0.50 = $15.00
- Total: $37.50 per month
Document summaries
5,000 documents per month, each 20,000 input tokens, with 500-token summaries.
- Input: 5,000 x 20,000 = 100M tokens x $0.15 = $15.00
- Output: 5,000 x 500 = 2.5M tokens x $0.50 = $1.25
- Total: $16.25 per month
Long-form writing
2,000 requests per month, each with a 500-token brief and a 3,000-token result.
- Input: 2,000 x 500 = 1M tokens x $0.15 = $0.15
- Output: 2,000 x 3,000 = 6M tokens x $0.50 = $3.00
- Total: $3.15 per month
The pattern: input-heavy jobs such as summarization are the cheapest per useful result here, and output-heavy jobs cost relatively more because of the higher output rate. When you compare with another provider, check how their input to output ratio differs, since that shifts which workload looks cheaper.
For a quick sense of scale, a request with 2,000 input and 500 output tokens costs $0.00055, so 1,000 of them cost $0.55. The $0.50 trial credit covers about 900 such requests.
Where price is not the deciding factor
A lower price per token means little if the model cannot do your task, or if a needed feature is missing. Here is an honest list.
Choose OpenAI, or another large hosted provider, if
- You need image input or generation, audio, video, embeddings or fine-tuning. This service is text only and has none of those.
- Your task needs the strongest available reasoning on hard problems. Test our model on your own prompts; do not assume any open-weight model matches a frontier model on every task.
- You need a vendor with an enterprise agreement, a signed data processing contract or a published uptime commitment. I make no uptime claims here.
- Your product depends on a specific vendor's model name.
Choose this service if
- You want a model that answers without content refusals for lawful adult use: fiction, roleplay and companion apps for adults, red-team text, unfiltered drafting. One hard limit applies: no sexual content involving minors.
- You want plain pay-as-you-go billing, with credit that never expires and failed or refused requests not charged.
- You already use the OpenAI SDK and want to change only the base URL and key.
- You want crypto payment with no subscription.
Many teams use more than one provider: a frontier model for the hard cases and a cheaper, less restricted model for volume. Since the API is OpenAI-compatible, switching between them is a configuration change.
Switching takes two lines
If your code already uses an OpenAI SDK, trying this endpoint means changing the base URL, the key and the model name. Nothing else in your request code has to change for plain chat completions, streaming, tool calling or JSON mode.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.uncensoreddeepseek.com/v1",
apiKey: process.env.API_KEY,
});
const res = await client.chat.completions.create({
model: "uncensored",
messages: [{ role: "user", content: "Give me three names for a fantasy tavern." }],
});
console.log(res.choices[0].message.content);
Limits to plan for: a full key allows 8 concurrent requests and 300 requests per minute, and a trial key allows 2 and 60. Past a limit you get HTTP 429 with the code rate_limited or concurrency. The body limit is 8 MB. The docs list the supported parameters: temperature, top_p, stop, seed and the penalties are passed through.
Because the setup is cheap to try, the practical advice is to run your own prompts on both sides. Sign up, claim the $0.50 trial credit on the API key page (one per person, 7 days, no card), and compare the answers and the costs from the script above.
A checklist for your own comparison
- Collect 50 to 100 real prompts from your product and note typical input and output sizes.
- Run them against each candidate and log token usage.
- Apply the formula with each provider's current prices, including any fees outside per-token billing.
- Read the answers side by side and mark which ones are usable without editing.
- Divide cost by the number of usable answers. That is the figure that matters, not the price per million tokens.
A cheap model that needs three tries per usable answer is not cheap. A premium model that nails it the first time may be. Only your own prompts can tell you which case you are in.
More guides
Questions and answers
Is your API cheaper than OpenAI?
It depends on the model you would use there and your input to output mix. Our rates are $0.15 per 1M input and $0.50 per 1M output; check OpenAI's current pricing and run the formula with your own token counts.
Is this the official DeepSeek API?
No. It is an independent OpenAI-compatible service running one open-weight model under the ID uncensored. For DeepSeek's own pricing, check their site.
Are there subscription fees or minimums beyond top-ups?
No subscription. You pay from prepaid credit that never expires, with top-ups from $20 to $500 in any whole amount.
Do failed requests cost money?
Failed or refused requests are not charged. Only successful completions are billed per token.
Your key is one form away
Create an account, copy the key, change the base URL. That is the whole setup.